The modeling of abnormal behavior in ‘normal’ subjects (often animals) has a long history in pharmacological research for the screening of novel drug compounds. Systematic criteria have been outlined in that literature to estimate the external validity of a model, that is to estimate how closely the model is linked to the disorder of interest. Experimental psychopathology (EPP) also uses behavioral models to study the psychological processes that underlie abnormal behavior. Although EPP researchers may occasionally feel uneasy about the validity of the model that they use, the issue has not received direct attention in this literature. Here, we review the criteria of validity as set out in pharmacology research (face, predictive and construct validity) and discuss their relevance for EPP research. Furthermore, we propose diagnostic validity as an additional criterion of external validity that is relevant to EPP research. We evaluate two models for the study of anxiety and depression, and show that they have good face, diagnostic and construct validity. However, EPP research generally lacks direct tests of predictive validity. We conclude that combined evaluations of predictive, diagnostic and construct validity provide a sound basis to infer the external validity of behavioral models in EPP research.